Skip to content

Fix Stable LM 3B build #1061

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 14, 2023
Merged

Fix Stable LM 3B build #1061

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5dc1bb3
[stablelm 3b] Rename dynamic vocab size from "v" to "vocab_size"
jeethu Oct 13, 2023
b856b81
Add get_num_key_value_heads method to StableLM3bConfig
jeethu Oct 13, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 11 additions & 4 deletions mlc_llm/relax_model/stablelm_3b.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,6 +66,11 @@ def __init__(
self.num_shards = 1
self.kwargs = kwargs

def get_num_key_value_heads(self):
if self.num_key_value_heads is None:
return self.num_attention_heads
return self.num_key_value_heads


class LayerNorm(nn.Module):
def __init__(
Expand Down Expand Up @@ -579,7 +584,7 @@ def create_embed_func(
bsz = 1
seq_len = tvm.tir.Var("n", "int64")
with bb.function(func_name):
model = StableLM3bEmbedTokensWrapper(config, tvm.tir.Var("v", "int64"))
model = StableLM3bEmbedTokensWrapper(config, tvm.tir.Var("vocab_size", "int64"))
param_manager.register_params(model, func_name, quant_scheme, get_param_quant_kind)

input_ids = nn.Placeholder((bsz, seq_len), dtype="int32", name="input_ids")
Expand Down Expand Up @@ -608,7 +613,7 @@ def create_encoding_func(
all_seq_len = tvm.tir.Var("m", "int64")
hidden_size = config.hidden_size
with bb.function(func_name):
model = StableLM3bForCausalLM(config, tvm.tir.Var("v", "int64"), sep_embed)
model = StableLM3bForCausalLM(config, tvm.tir.Var("vocab_size", "int64"), sep_embed)
param_manager.register_params(model, func_name, quant_scheme, get_param_quant_kind)

inputs = (
Expand Down Expand Up @@ -652,7 +657,7 @@ def create_decoding_func(
all_seq_len = tvm.tir.Var("n", "int64")

with bb.function(func_name):
model = StableLM3bForCausalLM(config, tvm.tir.Var("v", "int64"))
model = StableLM3bForCausalLM(config, tvm.tir.Var("vocab_size", "int64"))
param_manager.register_params(model, func_name, quant_scheme, get_param_quant_kind)

input_ids = nn.Placeholder((bsz, 1), dtype="int32", name="input_ids")
Expand Down Expand Up @@ -714,7 +719,9 @@ def create_kv_cache_func(bb: relax.BlockBuilder, config: StableLM3bConfig) -> No

def create_softmax_func(bb: relax.BlockBuilder, config: StableLM3bConfig) -> None:
with bb.function("softmax_with_temperature"):
logits = nn.Placeholder((1, 1, tvm.tir.Var("v", "int64")), dtype="float32", name="logits")
logits = nn.Placeholder(
(1, 1, tvm.tir.Var("vocab_size", "int64")), dtype="float32", name="logits"
)
temperature = nn.Placeholder((), dtype="float32", name="temperature")
with bb.dataflow():
div = bb.emit(relax.op.divide(logits, temperature))
Expand Down